Background
Determining and understanding predictive factors of postoperative analgesic consumption may help to anticipate patient’s needs and prescribe well-targeted analgesia. Several authors studied predictive factors of the severity of postoperative pain and highlighted the following elements: younger age, female gender, type of surgery, incision length, quantitative sensory testing, severity of preoperative pain, use of analgesics before surgery, psychological background, and genetic characteristics [
1‐
3]. In that endeavour, little attention has been paid on the influence of two aspects related to postoperative analgesics requirements, i.e. the individual’s attitude towards painful situations and the status of his immune system. In this study, we implemented a new questionnaire, the Situational Pain Scale (SPS), to measure an individual’s attitude towards imaginary potential painful situations. This questionnaire was calibrated with the one-parameter logistic Rasch model [
4] and designed to be invariant for age and gender. We hypothesized that patient’s postoperative analgesic requirements may be related to his attitude towards painful situations and that the status of his immune system plays an important role in mediating this relationship.
To assess the status of the patient’s immune system, we used the neutrophil-to-lymphocyte ratio (NLR) that has been proposed as one of the most sensitive markers to stratify patients in terms of inflammation [
5‐
9].
The chief aim of this study was to evaluate the predictive value of these two variables, in isolation and in combination with other variables, for the analgesic requirements during 48 h following laparoscopic cholecystectomy.
Results
All the 60 patients completed the study and no missing values were in the collected data.
Population characteristics
The population characteristics (age, gender, SPS, NLR, HADS, VRS, surgery duration) as analgesics requirements are reported in Table
1. Although the group of males was on average older (62.8 years) than the females (54.7 years) the mean difference of 8.2 years did not reach statistical significance (F = 2.884; P = 0.095). The mean (±SD) scores for anxiety and depression on the HADS test were 4.9 ± 3.9 and 9.1 ± 4.2, respectively. There was no statistical difference in mood between genders (F = 0.052; P > 0.5). Thirty-three (55 %) patients didn’t receive any opioids during surgery, but some of these patients received opioids in the recovery room so that finally only 8 patients were not exposed to opioids before returning to their room (Table
1).
Pre-surgical patients and the situational pain scale
The distribution of measures for the attitude towards painful imaginary situations (SPS) for the pre-surgical patients is shown in Fig.
1 (third panel). These measures range from approximately −4.5 to 3 Logits, with higher values associated with higher pain intensity reports.
Univariate analyses of potential predictors for postoperative analgesic consumption
The univariate analysis (Table
1) revealed a significant association between postoperative analgesic consumption (UD) and the following variables: SPS, total HADS scores, preoperative neutrophil counts, lymphocyte counts and the NLR. A negative relationship between SPS and the NLR (Ln(NLR) = 0.820–0.180*SPS; R
2 = 0.211; F = 15.52;
P < 0.001) and a positive relationship between SPS and the total score on the HADS (HADS = 14.8 + 1.63*SPS; R
2 = 0.159; F = 10.97;
P = 0.002) was observed. The logarithmic transformation of NLR was used as it markedly reduced the positively skewed distribution.
No significant associations were observed between UD and gender, age, preoperative pain intensity at rest and during movement measured by the VRS at rest and during movements, surgery duration and opioid administration (intraoperative period and in the Post Anaesthesia Care Unit).
Multivariate analysis of analgesic consumption (primary endpoint) and maximum pain on the VRS (secondary endpoint) in the postoperative period
A multiple linear regression analysis based on the variables identified as significantly related to postoperative analgesic consumption yielded the following results:
UD = 3.18 + 0.10 * SPS − 1.24 * NLR + 0.10 * HADS; R2adj = 0.253; F3,56 = 6.309; P < 0.001. The contribution of NLR to the determination of UD was clearly the most effective.
The same analysis performed with the maximum pain on the VRS during the same postoperative period showed that this relation was clearly weaker as only 9.5 % of the variance in pain intensity was explained by the independent variables:
$$ \mathrm{VRSmax} = 1.74 + 0.08*\mathrm{S}\mathrm{P}\mathrm{S}\hbox{--} 0.24*\mathrm{N}\mathrm{L}\mathrm{R} + 0.02*\mathrm{HADS};\ {\mathrm{R}}^2\mathrm{a}\mathrm{d}\mathrm{j} = 0.095;\ {\mathrm{F}}_{3,56} = 3.074;P = 0.035 $$
In the following section, we present a statistical mediation analysis for examining more precisely the relationship among the predictor variables of postoperative analgesic consumption.
We followed Baron and Kenny’s steps [for a comprehensive review see [
22]] for examining mediation in the present set of variables with the two-mediator model illustrated in Fig.
2. First step, the independent variable X (SPS) must affect the dependent variable Y (UD), i.e. the correlation coefficient c in the upper part of Fig.
2. That relationship was indeed significant (R = 0.321:
P = 0.012). Second step, the independent variable X (SPS) must affect the first mediator (M
1 or NLR) coefficient a
1 and must affect the second mediator M
2 or HADS) coefficient a
2. These effects were both highly significant (see Table
2) implying that both variables were mediators of the relation between SPS and UD. Third, the mediator must affect the dependent variable (UD) when the independent variable (SPS) is controlled: coefficient b
1 for the first mediator (NLR) and b
2 for the second mediator (HADS). For both mediators the coefficients are significant with
P = 0.017 and
P = 0.033, respectively. Fourth and finally, the direct effect c’ (lower part of Fig.
2) must be no significant. Consequently, there was clear evidence for a complete mediation since the direct effect was no significant (
P > 0.5) but a
1*b
1 was significant (
P = 0.005) although a
2*b
2 (
P = 0.161) was not.
Table 2
Paths coefficients and statistics of the two-mediator model (see Fig.
2 lower panel)
Symbols | a1
| a2
| b1
| b2
| c' | a1 x b1
| a2 x b2
| (a1 x b1)+ (a2 x b2)
| m2 < − > m1 | Contrasta
|
Coefficient | −0.180 | 1.634 | −1.243 | 0.102 | 0.097 | 0.224 | 0.166 | 0.390 | −0.093 | 0.057 |
S.E. | 0.038 | 0.247 | 0.365 | 0.063 | 0.361 | 0.079 | 0.119 | 0.196 | 0.435 | 0.047 |
t statistic | −4.774 | 6.612 | −3.403 | 1.608 | 0.269 | 2.819 | 1.401 | 1.985 | −0.214 | 1.219 |
P value | < 0.001 | < 0.001 | 0.001 | 0.108 | 0.788 | 0.005 | 0.161 | 0.047 | 0.831 | 0.223 |
Complete results of the analysis are reported in Table
2. As already mentioned here above, the need for analgesics (UD) was significantly related to the attitude towards imaginary painful situations (SPS) (c = 0.49; R
2 = 0.103; F = 6.68;
P = 0.012). In other words, a 1 unit increase in the SPS was associated with about half a unit increase in UD. This total effect can be explained by the mediated effects through the general state of inflammation measured by the NLR and mood measured with the HADS. There was a statistically significant effect of SPS score on NLR (a
1 = −0.180; R
2 = 0.211; F = 15.52;
P < 0.001) and on the HADS (a
2 = 1.634; R
2 = 0.159; F = 10.97;
P = 0.002). SPS was associated with a reduction of −0,18 in the NLR mediator and 1.63 change in the HADS mediator. The effect of the NLR mediator (b
1 = −1.243; F = 6.09;
P = 0.017) and the HADS mediator (b
2 = 0.102; F = 4.76;
P = 0.033) on UD was statistically significant when controlling for SPS. A 1 unit change in the NLR mediator was associated with a −1.24 decrease in UD and a 1 unit increase in HADS was associated with a 0.10 increase in UD. The adjusted effect of SPS on UD was not statistically significant (c’ = 0.097; F = 0.21;
P > 0.5) consistent with a random association of SPS and UD during the 48 h following surgery. Apparently the overall significant relation between SPS and UD was due too the effects of SPS on the mediators. There was a drop in the value of c’ (=0.097) compared with c (=0.487) of 0.390.
The estimates of the two mediated effects were equal to a1xb1 = 0.224 for mediation through NLR and a2b2 = 0.166 for the mediation through HADS. The total mediated effect of a1b1 plus a2xb2 = 0.390, which is equal to c–c’ = 0.487-0.097, so that a 1 unit increase in attitude was associated with a 0.39 effect on UD through the two mediating variables. In other words, the variance in UD explained by the score on the SPS was indirect and amounts to 46 % through mediator NLR and to 34 % through mediator HADS. The total mediated effect explained 80 % of the variance in UD.
Finally, there was no significant interaction between the two mediators (t = 1.219; P = 0.223).
Discussion
The present study shows that preoperative pain-related attitudes, as assessed by the SPS, were associated with the postoperative analgesic requirements (UD). Moreover, a higher score of anxiety and depression (HADS), or a lower preoperative NLR (and its components: a low neutrophil or a high lymphocyte counts) were also associated with a higher UD. Finally, and importantly, the HADS and the NLR mediate conjointly the effect of the SPS on the UD. In other words, patients with the strongest scores on the SPS, i.e. those who anticipate the highest pain scores, showed the lowest inflammatory status (assessed by the NLR) and the worst mood status (assessed by the HADS) that mediate, at least in this series, the effect on postoperative analgesic consumption (UD).
Others studied predictive factors of severe postoperative pain. For example, Kalkman et al.[
1] developed a prediction tool for the risk of early severe postoperative pain. They found that young age, female gender, outpatient, high preoperative pain score, anxiety and need for information, type of surgery and large incision size, all are predictive of severe postoperative pain. In their work, laparoscopic cholecystectomy was classified as a surgical procedure with “highest expected pain” [
1]. More recently[
23], preoperative Quantitative Sensory Testing (QST) has been shown to be better correlated with postoperative pain, than demographic and psychological factors like vulnerability, anxiety, depression, catastrophizing. Indeed, responses to experimental (thermal or electrical) pain stimuli explain up to 54 % of the variance of clinical postoperative pain. Adding the psychological variables to the multivariate regression analysis didn’t significantly increase the predictive power of the model. This led to hypothesize that there is multicollinearity between psychological and sensory variables. In the present study, we confirm and quantify multicollinearity between psychological variables and inflammatory response components. Nevertheless, our primary endpoint was here original, the analgesic needs (UD), in comparison with pain scores and/or the risk of severe postoperative pain, as described in the previous studies. Our analysis concerning the pain scores shows that, if the same relations cannot be excluded, these are clearly weaker than with the UD. Therefore, we identified here a potentially interesting new behavioural variable, associated with postoperative pain, but more sensitive in the context of the analysis of multicollinearity between psychological variables, inflammatory response and pain behaviour.
This multicollinearity comes not as a surprise as pain-related behaviour was linked to lymphocytes and neutrophil counts in psychological intervention on depressive symptoms in cancer patients. Indeed, using a mediation analysis, as in the present study, Thornton et al. [
24] showed that the effect of psychological intervention on white blood cell count in breast cancer patients was mediated by the reduction of depressive symptoms. This effect was mostly apparent on a reduction of the neutrophils count, but also present on lymphocytes count. They hypothesized that psychological intervention was associated with a shift from a constantly over activate immune response to a more, and better responding, adaptive immune status. In that case, immune response should be characterized by lower lymphocytes and neutrophils counts in the basal state but an increased inflammatory response resolving rapidly during the postoperative phase.
Using preoperative relaxation technique, Manyande et al. [
25] reported that a reduction of the anxiety before and after surgery induced a more robust cortisol and epinephrine response (typically associated with a stronger acute physiological stress response). Interestingly, in their study, postoperative pain scores were similar between the groups (relaxation or not) whereas the analgesic requirements were twice less in patients with relaxation intervention, but with stronger inflammatory response. We observed similarly an association between psychological variables, inflammatory response and postoperative analgesics requirements.
The increased preoperative inflammatory status we measured is concordant with the association seen between a high redistribution profile after an inflammatory event (i.e. high neutrophils counts and rapidly decreasing lymphocytes count after surgery) and a better recovery and functional status [
26]. We can hypothesise that the patients with a preoperative low SPS score could have a more responsive adaptive immune response profile but this remains speculative and should be tested in further studies.
The exploratory way used is the main limitation of this work. The use of the SPS remains a new tool to assess attitude towards painful situations, as the endpoint proposed to assess the analgesics requirements (total UD during the postoperative period)(For details over procedure and SPS calibration, see
Appendix: Table 3, Additional files
1 and
2). Nevertheless, the correlation between the SPS and the UD permits to describe logical associations in the absence of satisfactory alternative tools (for these types of status and behaviour assessments). Additionnally, regarding the difference between UD and analgesic requirements, one can argue that analgesics requirements are multifactorial and the number of analgesic requirements depends primarily on the type, the pharmacokinetics and pharmacodynamics of the analgesic(s) used. This is not considered primarly in this work, defining "patients' requests" for analgesia as primary endpoint. We consider as strengths the use of the NLR as a sensitive marker of the inflammatory status, the wide use of the HADS for the assessment of the emotional status, as the clear results obtained by the mediation analysis. Finally, we recognize that the adjusted comparisons on patient’s weight, while malnutrition and morbid obesity were exclusion criteria, would merit additional comparisons.
Competing interest
The authors attest no conflict of interest.
Authors’ contributions
MD : Study design, subjects recruitment, data collection, manuscript preparation and final acceptance; CD: Study design, subject recruitment, data collection, data analysis, manuscript preparation and final acceptance; BlPdW: Study design, manuscript preparation and final acceptance; JLT: Study design, manuscript preparation and final acceptance; LP: Study design, data analysis, manuscript preparation and final acceptance; PF: Study design, subject recruitment, data collection, data analysis, manuscript preparation and final acceptance. All authors read and approved the final manuscript.